Transportation Research Record 1022 7

12. E. Kausel and R. Peek. Dynamic Loads in the
Interior of a Layered Stratum: An Explicit So-
lution. Bulletin of the Seismological Society
of America, Vol. 72, No. 5, Oct. 1982, pp.
1459-1481.
13. R.J. Apsel. Dynamic Green's Functions for Lay-
ered Media and Applications to Boundary-Value
Problems. Ph.D. dissertation. University of
California at San Diego, 1979.
14. F.E. Richart, Jr., J .R. Hall, Jr., and R.D.
Woods. Vibrations of Soils and Foundations.
Prentice-Hall, Inc., Englewood Cliffs, N.J.,
1970.
Publication of this paper sponsored by Committee on
Mechanics of Earth Masses and Layered Systems.

Dynamic Interpretation of Dynaflect and

Falling Weight Deflectometer Tests
JOSE M. ROESSET and KO-YOUNG SHAO

ABSTRACT

The Dynaf lect and the falling weight def lectometer are commonly used for non-
destructive testing of pavements. In both cases a dynamic load is imparted, and
the determination of the mechanical properties of the pavement, the base, and
the subbase is normally performed by comparing the measured deflections at var-
ious points along the surface to results of static analyses that consider the
subbase as a homogeneous, elastic half-space. In this paper, the displacements
obtained from dynamic analyses are compared to those provided by conventional
static programs when the subbase is a homogeneous soil stratum of finite depth
resting on a much stiffer rocklike material and when the soil properties
increase smoothly with depth, as is often the case. The results of these
comparisons indicate that for certain ranges of depth to bedrock a static
interpretation of the Dynaflect and falling weight deflectometer tests may lead
to substantial errors. Situations in which these errors are important are more
likely to be encountered with the Dynaflect than with the falling weight de-
flectometer.

The Dynaflect and the falling weight deflectometer
are commonly used for nondestructive testing of
pavements. The Dynaflect consists of a force genera-
tor and five geophones housed in a small trailer,
which is towed by a light vehicle. The loading sys-
tem consists of two counterrotating eccentric
masses. The resulting vertical force varies harmon-
ically with time. At a frequency of B Hz, a 1,000-lb
peak-to-peak oscillating force and a base load of
1,000 lb are transmitted to the pavement through the
loading wheels. The resulting deflection basin is
measured by five geophones that are mounted on the
trailer draw bar at 12-in. intervals. The positions
of the geophones (STl through STS) with respect to
the wheels are shown schematically in Figure 1.
The falling weight deflectometer has a 330. 7-lb
(150-kg) weight mounted on a vertical shaft and
housed in a compact trailer that can easily be towed
by most conventional passenger cars. The weight is
hydraulically lifted to a predetermined height
(ranging from 0 to 15. 7 in. or 0 to 400 mm). I t is
then dropped onto a rubber pad 11. 8 in. ( 300 mm)
thick, that helps to distribute the load uniformly
over the loading area. The resulting load is a force
impulse with a duration of approximately 30 msec and
a peak magnitude ranging from 9 to 14,000 lbs (0 to
60 000 N) depending on the drop height. The peak
force and maximum deflections at various points
along the surface are measured by load cells and
velocity transducers. The applied pressure is mea-
sured in kilopascals and the deflections in microme-
ters.
In the case of the Dynaflect the deflections mea-
sured at the various stations represent the ampli-
tudes of the steady-state displacements at a given
frequency ( 8 Hz) For the falling weight deflectom-
eter they are the peak displacements under a tran-
sient-type excitation. In both cases the tests are
dynamic in nature, but the interpretation of their
results to estimate the elastic properties of the
pavement, base, and subbase relies on static analy-
ses. Furthermore, these analyses assume that the
soil in the subbase is an elastic, homogeneous, and
isotropic half-space. In many cases soil properties
8 Transportation Research Record 1022
DYNAFLECT
P=500 lb,F=SHz
.,,.
I tained for each term of the series (each frequency)
:1 2 3 4 5
I
I For an isolated layer wi th uniform properties,
I
)le Scale: I ft
~ bottom surfaces can be expanded in a double Fourier
P=500 lb
FWD
P(I)
>"'1
! I~ Q
3 4
Q Q
5
0
6
II
7
R the top (or vice versa). This approach [Thomson (1)
1000 lbt-/\ Scale'~
O~t
0 0.033
FIGURE 1 Geometric configuration of loads and
stations for Dynaflect and falling weight
deflectometer,
will vary with depth a nd the soil will be underlain
at some depth by stiffer, rocklike material.
The purpo$e of this work is to determine the dy-
namic displacements at points along the surface of a
pavement excited by forces simulating the excitation
of the Dynaflect and the falling weight deflectom-
eter. These displacements are compared for various
depths to bedrock with those resulting from static
analyses for the same soil profile and assuming an
elastic half-space (the normal assumption). The dy-
namic deflection bulbs obtained from the analyses
are then used as inpu t for the standard backfiguring
process to estimate the elastic moduli of the pave-
ment, base, and subbase in order to assess the er-
ror s induced by neg.lecting dynamic effects.
FORMULATION
Consider a soil deposit that consists of horizontal
layers. The mass density and the elastic moduli of
the soil may change with depth, from layer to layer,
but are assumed to be constant over each layer. For
the present application the top layer would repre-
sent the pavement (assuming that it extends to in-
finity in both horizontal directions), the second
layer would be the base, and the remaining layers
would represent the soil of the subbase. An accurate
solution would require consideration of the finite
width of the pavement. Even so, for the purposes of
this study, these simplifying assumptions should not
be unreasonable . The determination of the response
of this soil deposit to dynamic loads applied at the
surface (or at any point within the s oil mass) falls
mathematically into the area of wave propagation
theory.
The formulation of these problems always starts
by considering steady-state harmonic forces and dis-
placements at a given frequency. For a harmonic ex-
citation, as ca used by a vibrat ing machine rotating
at a specified velocit.y (case of the Dynaflect), the
solution at the corresponding frequency provides
directly the desired results. For an arbitrary
transient excitation (case o f the falling weight
deflectometer) , the time history of the specified
forces must be decomposed into different frequency
components using a Fourier series, or more conve-
niently a Fourier transform. Results are then ob-
and combined to obtain the time history of displace-
ments (inverse Fourier transform).
the stresses and displacements along the top and
series (or Fourier transform) in the two horizontal
d ireotions f or Cartesian coordinates, or in a Fou-
rier sei:ies in the circumferential direction and a
series of modified Bessel functions in the radial
direction for cy lindr ical coordinates. For each term
of these series, corresponding to a given wave num-
ber, there can be determined closed-form analytical
expressions in the form of a transfer matrix relat-
ing amplitudes of stresses and displacements at the
bottom surface to t he corresponding quantities at
and Haskell ( 21 I has served as the basis for most
s tudies on wave propagation th.rough layered media in
the last 30 years . An alternative is to relate the
stresses at both surfaces to the displacements, ob-
taining a dynamic stiffness ma tr ix for the layer
(3), which can be used and understood in much the
same way as those in structural analysis. For a
half-space, the s t iffness matrix relates directly
stresses and displacements at the top surface be-
cause tbe bottom surface is pushed to infinity .
Assembling the stiffness matrices of the different
layers, there oan be obtained a stiffness matrix for
the complete soil deposit, which relates forces per
unit of area applied at the free surface, or the
interfaces between the layers, to the displacements
at the same elevations.
The terms of the transfer or stiffness matrices
of each layer are transcendental functions (complex
exponentials). In addition, results must be obtained
for. each term of the Yourier series decomposition
(each wave number), then combined, normally by
numerical integration, to obtain the solution for a
specified load distribution. On the other hand, the
thickness of the layers is controlled only by physi-
cal considerations and the assumption of uniform
properties. This makes the procedure particularly
convenient when dealing with a homogeneous half-
space or a small number of layers but extremely ex-
pensive when a large number of layers are needed to
reproduce properly the variation o f soil properties
with depth. Formulations along these lines have heen
implemented by Gazetas (!J in Cartesian coordinates
and by Apsel C21 in cylindrical coordinates.
When the layers are extremely thin, the transcen-
dental functions representing the variation of dis-
placements with depth can be approximated over each
layer by a straight line (or higher order polynomial
expansions). The solution (displacements and
stresses) is then expressed in terms of the exact
a nalytical expressions i n the two hori zontal (or
radial and circumferential) directions and in terms
of simpler polynomial. expansions in the vertical
direction (as in a finite element formulation). This
approximation leads to much simpl.er algebraic ex-
pressions for the terms of the transfer or stiffness
matrices of tile layers. In addition, when the soil
is underlain by a much stiffer, rocklike material,
which can be considered rigid, i t is possible to
determine the wave numbers (eigenva.lues) and the
mode. shapes (eigenvectors) of the waves propagating
through the soil deposit by solving an algebraic
eigenvalue problem (6 1 7). Expressing the solution in
terms of these mod-; - shapes (eigenfunction expan-
sion), Kausel (Bl was able to obtain explicit solu-
tions for the displacemente caused by harmonic dy-
Roesset and Shao
namic l oads i n a l aye r e d soil deposit. Kausel's
formulation is part icula rly efficient from the point
of v i ew of computation, but the layers must be suf-
f i ciently t h i n t o rep roduce accurately the variation
o f the d i splace me nt s wi th depth with a piecewise
l i near appr ox i ma tion.
Because the p u r pose of t his .wor k wa s to i nv es t i-
gat e the e f f ects o f dep t h to bedro ck and variation
o f soil p r ope r ti e s wi t h d ept h o n the dynamic r e -
sponse of a p avemen t , it was decided t o use Kause l' s
f ormulat i on. Th e f ormu lation was impleme n ted i n a
computer prog r.am a nd t h e r esults were c ompared with
those p ublis hed by Ka use l (]_) with e xcellen t ag r ee-
ment . Because of the disc r ete nature o f the f o r mu l a-
tion, b e fore it i s a pplie d, an approp r ia t e mes h si ze
(thickness of the sublayers) to guarantee an ac-
curate solution must be determined.
Studies were conducted first for static loads
(zero frequency), a homogeneous soil deposit of
f i n i te qep t h , a nd a vertical load on the surface
unifor mly d i s tributed ove r a cicoula r area with a
radi us (r0 ) o f l in. (s i mul ati ng the loading in
the Dy naflect) o r 6 i n. (approx imate d imensions of
the loadi ng pla te of t he falli ng weigh t deflectom-
eter). The properties of the soil deposit are shown
in Figure 1. This represents another approximation
because the load distribution for the Dynaflect will
be more nea r l y e l liptic a l. This simp lifica tion ap-
pears to be justified f or the p urposes o f t h is
study. A model wi th all layers of t he same th i ck ne s s
was init ially conside r e d. Figure 2 shows typ ical
results for a deposit 40 ft deep. The displacements
at the center of the loaded area and at distances
(d) of 2 and 4 ft from this point are divided by the
exact solution and plotted versus the inverse of the
number of layers. Ten layers correspond, therefore,
to a l ayer t hic kness of 4 ft . :tt can be seen f rom
t his figur e that e xce lle n t res ults a r e obtai ned at
di s tances o f 2 a nd 4 f t e ven with th e c o arser mesh
(.10 l ayers ) Th e erro r for a d i stance o f 2 ft is
s lightly l ar g er f or the s mal l l oa ded a r ea ('radius of
l in.) , bu t i t i s only 2 pe rce nt wi t h t h e coa r se
mesh. Results at the center of the loaded area are,
on the other hand, extremely poor even when taking
H 40 ft
OO-l-~---.---~~~~~~-"T"""---;-
_I _\_ -1 _L
40 20 10 N and so forth. When results are desired under the
FIGURE 2 Variation of displacements with number of
layers at H = 40 ft, WE =exact solution.
9
40 layers (thickness of l ft for each layer), and
they are much worse for the small loaded area. This
indicates that for static loads the thickness of the
layers has to decrease with decreasing distance be-
tween the load and the point where displacements are
computed.
Figure 3 shows the ratio (Wff/~) where Wff is the
displacement for a stratum of depth H and w.., is
the displacement for a half-space. The displacements
are computed at the center of the loaded area and at
distances (d) of 2 and 4 ft. They are plotted versus
the inverse of the stratum thickness to better il-
lustrate the convergence rate. The results indicate
that the displacement is nearly inversely propor-
tional to H. It is interesting to note that at the
center of the loaded area the displacements for a
stratum with a depth of only 8 ft are already within
l percent of the results for a half-space with a
radius of l in. and within 5 percent for the 6-in.
radius. The depth needed to reproduce a half-space
increases clearly with increasing distance between
the load and the point where displacements are com-
puted. This suggests a l so t hat c l o s e to the load the
static displacements a re affe cted only by the soil
properties near the s ur fa ce, wnerea s f or increasing
distances the soil pr o per t ies a t large r depths will
influence the results more significantly.
\.2
WH 1.0
Wro
ro = 6 11
,CENTER POINT
0 .8
0 .6
0.4
0.2
FIGURE 3 Variation of displacements with depth to
bedrock.
On the basis of the observations from these two
series of studies, it was decided that an improve-
ment in the accuracy and the economy of the computa-
tions could be obtained by taking thin sublayers
near the surface and gradually increasing their
thickness with depth. The distance from the center
of the loaded area to the point where the displace-
ments are computed was designated o, and a rule was
derived to automatically generate a desired mesh.
According to this rule the first n ft are divided
into 2N sublayers of equal thickness, and the next D
ft are divided into N sublayers. N sublayers are
then used for the following 20 ft, the next 4D ft,
loaded area the distance o is replaced by the radius
of the loaded area. For a nonhomogeneous soil de-
posit (such as a pavement), the thickness of each
10 Transportation Research Record 1022

TABLE 1 Displacements for Different Meshes (10- 8 ft)

Radius Center Point d = 2 ft d = 4 ft

of Disk
(in.) H (ft) Fine Standard Coarse H (ft) Fine Stand ard Coarse H (ft) Fine Standard Coarse

6 2
8 35 .27 34.88 33 .4 8 8 2.9 7 1 2. 97 0 2.90 9 8 0. 7609 0.766 4 0.76 13
32 36 .58 36.20 34.78 32 4.237 4.230 4 .148 32 1.88 7 1. 884 1.84 7
128 3 6. 92 36.53 35. 10 128 4 .566 4.557 4 .472 128 2.21 4 2.209 2. 168
5.33 2 19.6 2 17 .3 208 .7 8 2. 93 2 2.9 30 2.87 3 8 0.7 55 3 0. 760 7 0. 7560
21.33 22 1.6 2 19.2 210 .7 32 4.200 4.1 92 4.11 3 32 1.88 3 1.8 79 1.84 3
85.33 222. 1 219 .8 2 1 1.2 128 4. 529 4 .5 19 4.437 128 2.209 2. 205 2.1 64
/.
sublayer is the smaller of the value suggested by
the rule or t .he actual physical dimension of the
layer. When the physical thickness controls, the MESH
mesh gene.rated according to the rule is subdivided t>- FINE
automatically to accommodate this cd terion. Fi-
nally, when the results are obtained simultaneously 0 o- STANDARD
at various points, the smallest D (or the radius of ~

the loaded area) controls. 0

This rule was used to construct meshes with
values of N equal to 4, 2, and 1. These will be
referred to as f i ne, stand.a rd, and coarse mesh, re-
spectively. Results were then obtained for a homoge-
neous soil deposit with a Young's modulus (E) of 20
-
w -,.,,.
- a
ksi and a Poisson' s rat i o of 0.4. The displacements bo
obtained with the coarse mesh are within 5 percent
of those of the fi ne mesh at the center of the c:
loaded area and improve in accuracy for greater dis- 0
tances. The results with the standard mesh differ :;:o
(.) C\I

f rom those with the fine mesh by less than 1.5 per-
cent at the center of the loaded area and are again
even closer for greater d i stances (Table 1). It was
concluded from these results that the standard mesh
s hould be sufficiently accurate for most practical
applications. Given the various approximations and
uncertai~ties involved in all phases of these analy-
ses, the coarse mesh may be adequate in rnany cases.
Using these three meshes and the same soil pro-
file, parametric studies were conducted next for a
dynamic excitation and different frequencies. It is
a commonly accepted rule of thumb, in dynamic stud-
ies using finite element models, that the she o f
the elements must be of the order of one-quarter to
one-sixth o .f the wavelength to obtain reasonably
accurate results. T.he wavelength is equal to the
shear wave velocity of the material divided by the
f requency for shear waves and the P wave velocity
divided by the frequency for compressional or dila-
tational waves. If E is Young's modulus of the ma-
terial, v its Poisson's ratio, and p its mass
density, the shear modu.lus is
G = E/2 (1 + v)
and the constrained modulus is
1 + 2G E(l - v)/(l + v) (l - 2v)
The shear wave velocity (vs) is then given hy
Vs
= G/p
and the P wave velocity is
v~ = (1 + 2G)/p = v2 2(1 - v)/(l - 2v)
The Rayleigh wave velocity, associated with surface
waves generated by a surface loading, is only
slightly smaller than the shear wave velocity (vs).
Figure 4 shows the amplitude of the steady-state
displacements obtained with the fine and standard
meshes at a point 5 ft from the center of the loaded
area. The displacements are plotted versus a dimen-
-
Q) 0
Q)
0
0
0
0-J-__,__,,____,__,.__,__,__..__,__,...----.__,.----.-l
0
00 0.40 1.20 2 .00
Dimensionless Frequency
FIGURE 4 Amplitude of displacements at Point 7,
fine and standard meshes.
sionless frequency. A value of the dimensionless
frequency of 1 corresponds to an actual frequency of
(1) 16 Hz and a wavelength of approx i mately 32 ft. The
maximum layer thickness, at the hottom of the soil
profile, in the standard mesh is 8 ft and in the
fine mesh 4 ft. It can be see.n that the res.u lts are
(2) in good agreement up to a dimensionless frequency of
about 1, corresponding to a wavelength equal to four
times the maximum layer thickness of the standard
mesh. For higher frequencies the results of the
(3) standard mesh exhibit a series of sharp peaks that
are not present in the more refined solution. Figure
5 shows similar results us ing the f i ne mesh and a
mesh with twl.ce the number of layers (each layer
(4) half the thickness of those in the fine mesh) , The
two solutions are almos t identical up to a t'limen-
sionless frequency of ahout 2, corresponding to a
wavelength equa l to four times the maximum layer
thickness of the f ine mesh. I n both cases the agree-
ment is even better when the displacements at closer
distances are cons idered. The d is tances involved in
the Dynaflect and f a ll.ing we i ght de fl.ectometer tests
are smaller than or equal t o 6 ft ,
Roesset and Shao 11

Figure 6 shows the results using the fine and

MESH coarse meshes. The differences in this case are of
the same order of magnitude as those reported for
o - FINE static loads in the low-frequency range, but they
1:> - REFINE become much more pronounced for dimensionless fre-
0
<:t quencies greater than 0. 4 (wavelengths less than 5
ci times the maximum layer thickness).
These results appear to confirm the validity of

-
the rule of thumb commonly used in practice. The
standard mesh will provide good results for the
....... 0 dynamic case as long as the wavelengths are longer
(!) - ~
than four times the maximum thickness of any layer.
For higher frequencies the mesh must be modified to
bo satisfy this additional constraint (reducing the
thickness of the bottom layers)
c
0 SIMULATION OF DYNAFLECT TESTS
+- 0
(.) C\J
Q) ci A pavement system was selected to evaluate the im-
....... portance of dynamic effects on the results of the
<!)
0 Dynaflect tests (Figure 7) The pavement has a
thickness of 2.5 in. and a Young's modulus of 200
ksi; the base has a thickness of 15 in. and a mod-
0 ulus of 78.5 ksi. The soil of the subbase was con-
ci sidered homogeneous with a Young's modulus of 29 ksi
and with a modulus starting with this value at the
top and increasing with depth. Different depths to
bedrock were used in the range from 10 to 110 ft.
Displacements were computed at the points corre-
sponding to the stations of the Dynaflect for a
o+-~~.--~~~~-.-~~...-~~.--~--r
0 static load and for a frequency of 8 Hz.
00 0.8 0 2.40 4 .00 Figure 8 shows the variation of the static dis-
Dimensionless Frequency placements with depth to bedrock at the five sta-
tions. Figure 9 shows the corresponding results for
FIGURE 5 Amplitude of displacements at Point 7, a frequency of 8 Hz, typical of Dynaflect tests.
fine and refined meshes.

:200000 Psi V= o.35"

MESH
-FINE
0 - COARSE
<:t
ci 1s
11 .J) =0.35

-
....... 0
(!) - ~
I 0
0

c
0
_o
(.) C\J
Q) ci
.......
Q)
0 E 2.9000 ;;s1.: ).) =- o.o/o
H
0
ci

o.+-~--.~~...-~--..~~.....--~-,-~--t
0
0 0 0.20 0.60 I. 00
Dimensionless Frequency
Roe. k
FIGURE 6 Amplitude of displacements at Point 7,
fine and coarse meshes. FIGURE 7 Profile of pavement used for studies.
12 Transportation Research Record 1022

6------------------.
0
.. 1.25
POINT I

POINT- I
wd
1.00
Ws 20 40 60 80 100 120
I/) SOIL DEPTH, ft
0
....... POINT-2 0 .75
._o
\.D 2.00
b E =CONSTANT
H=H
c POINT-3 for W5
01'0 1.75
:;:Q
uo
<V
POINT-4 1.50
......
<V
0
POINT-5 1.25 POINT 5
0
0
wd
1.00
o.._~~....-~---.~~--.-~~-.-~~....-~--f Ws 70 80 10
0
00 20.00 60.00 100.00
Soil Depth, H 0.75 SOIL DEPTH, ft
FIGURE 8 Variation of static displacements with
depth to bedrock-Dynaflect. 0 .50
FIGURE 10 Ratio of dynamic to static deflections, Points 1 and
5-Dynaflect.
b ~-----------------....
0
entirely. The range of depths over which there is a
POINT- I substantial dynamic amplification of the deflections
is closely associated with the depths for which a
I/) frequency of 8 Hz represents the natural frequencies
0 POINT-2 of the soil deposit in shear and dilatation. These
.::o
\.D
would be 20 and 48 ft, respectively.
Because the elastic properties of the pavement,
b base, and subbase are normally determined by compar-
ing the measured deflections to those resulting from
c POINT-3 static analyses assuming that the subbase is an
0 1'0
:;: Q elastic half-space, it is perhaps more interesting
uo to compare the dynamic results to the static deflec-
<V tions for an infinite depth to bedrock. The ratio of
POINT-4
...... these deflections for Points 1 and 5 is shown in
<V
0 Figure 11. These results indicate that for shallow
depths to bedrock (less than 20 or 25 ft) the dy-
POINT-5
0 namic deflections are smaller than the static de-
0 flections for a half-space (although they are larger
than the static deflections for the same soil pro-
0o.._~~.--~-,-~~--.-~~...-~---..--~-4
file with a finite depth). For a range of depths of
from 25 to 40 ft the dynamic results are larger than
00 20.00 60.00 100.00 the static ones because the dynamic amplification is
Soil D~pt h, H more pronounced as the distance to the load in-
FIGURE 9 Variation of dynamic displacements with creases. For depths greater than 50 or 60 ft the
ratio of dynamic to static displacements is close to
depth to bedrock-Dynaflect. l, It is thus for depths to bedrock of less than 40
ft that the errors committed by the present inter-
pretation procedures can be most serious for this
Figure 10 shows, finally, the ratio of the dynamic particular profile. (Greater depths would be signif-
to the static displacements at Points 1 (between two icant if the soil of the subbase were stiffer than
wheels) and 5 (farthest from the loads) As the the one selected for this study.)
depth to bedrock increases so does the ratio of dy- Determination of the characteristics of the pro-
namic to static deflections, reaching a peak for a file from the measured deflections falls into the
depth of approximately 35 ft and a second, much general category of system identification problems
sharper peak for a depth of about 42 ft and exhib- (sometimes referred to as the inverse problem), Be-
iting a sharp valley immediately after. As the depth cause only five deflections are available, it 'is
to bedrock continues to increase the ratio appears often assumed that the thickness of the pavement and
to tend to 1 from above. Additional studies assuming the base are known and that the only unknowns are
2 and 3 percent internal damping in the soil indi- the moduli of elasticity. These moduli are normally
cated that the first peak was only slightly affected estimated by a trial and error procedure, assuming a
by the existence of a small amount of material damp- set of values, computing the corresponding static
ing (which can always be expected) but that the sec- deflections, comparing them to the measured values,
ond peak and the following trough disappeared almost and iterating until the differences are smaller than
Roesset and Shao 13

1.25
wd 1.00
Ws 4 60 80
SOIL DEPTH,
0.75
2 .00
E =CONSTANT
H =INFINITE
1.75 for W5
1.50
1.25
Wct
1.00
Ws 0 70
0.75
0.50
FIGURE 11 Ratio of dynamic to static (H =~)displacements,
Points I and 5-Dynaflect.
an acceptable tolerance. Unfortunately, uniqueness
of the solution cannot be guaranteed and different
sets of elastic moduli can produce results that are
within the specified tolerance.
To get a better feeling for the significance of
the difference between static and dynamic displace-
ments, the deflection bulbs computed for depths to
bedrock of 10, 20, 35, and 110 ft were used as input
to the identification procedure. The exact values of
the elastic moduli were used as initial guesses and
a gradient search technique was used in an attempt
to converge to an optimum match using the computer
program BASSD2 (9). The results of these studies are
given in Table 2. Listed in the table are the com-
puted deflections, the estimated values of the elas-
tic moduli, and the errors in these moduli. It can
be seen that for a depth to bedrock of only 10 ft
the stiffness of the subbase is badly overestimated,
whereas the modulus of elasticity of the base as
well as the modulus of the pavement are underesti-
mated. This occurs because the dynamic and finite
POINT I layer effects are more pronounced for the farthest
stations, the deflections of which are heavily in-
fluenced by the soil properties at greater depths.
For a depth to bedrock of 20 ft the properties of
100 120 the base and the soil are accurately determined, but
ft the modulus of the pavement is badly overestimated.
For a depth of 35 ft the moduli of the pavement and
the base are both overestimated and the stiffness of
the subbase is underestimated. This situation is the
reverse of that encountered for a depth of 10 ft.
When the depth of bedrock is 110 ft the results are
more reasonable although the estimated modulus of
the pavement is still 24 percent too high.
It is important to keep in mind that these re-
sults are not unique and that another person might
obtain different values of the moduli with the same
quality of fit. Even so, it is believed that the
POINT 5 results illustrate reasonably well the type of er-
rors and the variation in estimated properties that
can be expected.
The same series of studies was conducted assuming
80 10 that the soil properties increased gradually with
depth. Figure 12 shows the ratio of the dynamic de-
flections for the soil profile with bedrock at a
finite depth to the static deflections assuming that
the subbase is homogeneous and extends to infinity.
Notice that in this case the range of depths over
which there is a substantial dynamic amplification
is somewhat larger (from 20 to 60 ft approximately)
because the subbase is effectively stiffer. An
amplification effect is still apparent for a depth
to bedrock of 110 ft whereas for the homogeneous
soil the ratio of dynamic to static deflections is
close to 1 for these depths.
SIMULATION OF FALLING WEIGHT DEFLECTOMETER TESTS
Because the loads applied by the falling weiqht de-
flectometer are transient in nature, it is neces-
sary, to simulate the results of this test, to de-
compose the time history of the force into frequency
components using the Fourier transform. Analyses
must then be conducted for a large number of dif-
ferent frequencies to obtain the transfer functions
of the deflections at each point (station). These
transfer functions are then multiplied by the Fou-
rier transform of the input and the resulting func-
tions are converted back to time using the inverse
Fourier transform. The final results are the time
histories of the deflections at the various points.
The complete analysis is clearly much more expensive
than is the case of the Dynaflect where only one
frequency is involved. Therefore the studies were

TABLE 2 Deflection Bulbs and Estimated Elastic Moduli for HomogeneouH Suhhase and
Different Depths to Bedrock-Dynaflect

Displacement (mils)
H Younfs Modulus Error
(ft) Point I Point 2 Point 3 Point 4 Point 5 (lb/in ) (%)

Static 0.70 0.52 0.36 0.26 0.20 200 ,000

78,500
29,000
Dynamic 10 0.61 0.44 0.28 0.17 0.11 150,000 25.0
30,000 61.8
45,000 55.2
Dynamic 20 0.68 0.51 0.35 0.24 0.18 340,500 70.3
78 ,000 0 .6
29,837 2.8
Dynamic 35 0.82 0.65 0.48 0.38 0,31 350,000 75 .0
98,500 25 .5
20,000 31.0
Dynamic 110 0.69 0.52 0.35 0.25 0.18 248 ,370 24.0
78 ,500 0.0
29,833 2.9
14
125 POINT
wd 1.00
40 60 80 100 120
Ws
SOIL DEPTH, ft
0 .75
2.00
E =VARIABLE
H =INFINITE
1.75 for W5
1.50
1.25 POINT 5
Wct
1.00
Ws 70 80 10
0.75
SOIL DEPTH, ft
0.50
FIGURE 12 Ratio of dynamic to static (H =~)displacements,
variable soil profile-Dynaflect.
limited to depths to bedrock of 10, 20, 40, and 80
ft.
The continuous Fourier transform involves an
integral over time (direct transform) or frequency
(inverse transform) extending from minus infinity or
zero to infinity. In practice, however, a discrete
transform, referred to as the fast Fourier trans-
form, is used. In this case a finite number of
points (power of 2) are selected to reproduce the
function of time at equal time intervals (6t). The
total duration is T ., Nt. t if N is the number of
points. Notice that for an impulse-type load the
values of the function will be nonzero for only a
few points. The Fourier transform is then calculated
at N/2 points with a frequency interval 6 f = l/T
and a maximum frequency (fmaxl equal to l/26t.
Proper selection of these parameters is important to
guarantee the accuracy of the final results. A small
time interval (tit) is desirable to reproduce prop-
erly the time variation of the forcing function and
to ensure that the peak response displacement is not
missed. The total duration (T) should be several
times larger than the actual duration of the load to
ensure that spurious free vibration terms have at-
tenuated 1 the appropriate value depends on the fun-
damental period of the system and the amount of
damping (in the present case no internal damping is
assumed for the soil and the only source of energy
dissipation results from radiation or geometric
spreading of the waves above the fundamental fre-
quency of the soil stratum). The frequency increment
llf [fixed when the duration (T) has been selected}
should be small to reproduce properly the transfer
function, particularly if it exhibits some sharp
peaks (typical of lightly damped systems). All these
considerations point out the desirability of a small
t.t and a large number of points N. It should be
noticed, however, that as the number of points in-
creases so do the cost of computation and the number
of frequencies for which analyses must be conducted.
As tit decreases, the maximum frequency ( fmax>
increases. This requires more refined meshes and a
Transportation Research Record 1022
larger number of layers because of the dynamic limi-
tation on the thickness of the layers.
A number of preliminary studies were conducted to
assess the values of At and N required to obtain
reasonably accurate results. It was concluded from
these studies that a value of N equal to 2 ,04 8 and a
time interval of approx imately 0. 002 sec were a p-
propriate for these applications. Figure 13 shows a
typical transfer for the center of the oaded area
and a depth to bedrock of 20 ft. (The transfer
function is actually complexi the amplitude of the
function is shown . ) It can be seen that for frequen-
c ies larger than 20 Hz the function is relatively
smooth without any pronounced peaks. It was decided,
therefore , to calculate the values of the transfer
functions at frequency intervals of approximately
0.25 Hz in the range from 0 to 20 Hz, 2 Hz from 20
to 60 Hz, and 4 Hz from 60 to 120 Hz. Because the
t.f required is of the order of 0.25 Hz the values
of the transfer functions at intermediate points are
evaluated by interpolation between the computed val-
ues. Finally, because fmax should be approximately
2 40 Hz the values between 120 and 2 40 Hz were ob-
tained by extrapolation. The preliminary studies
indicated that the results obtained with these sim-
plifications (leading to considerable savings in
computer time) were in good agreement with those
obtained using a constant frequency increment of
0.25 Hz over the complete range of freqencies.
Figure 14 shows typical time histories of the
displacements at Point 1 (center of the loaded area)
and Point 7 (farthest station) for a depth to bed-
rock of 20 ft. From these figures the peak deflec-
tion was computed at each station and the deflection
bulb was obtained. Figure 15 shows the ratio of the
dynamic to the static deflections considering both a
finite layer and a half-space for the static analy-
ses. It can be seen that a small amount of dynamic
amplification takes place particularly as the dis-
tance to the load increases although dynamic effects
are much less pronounced than in the case of the
Dynaflect.
0
0
0
rt)
0
0 +-..-..-..-.--~~--~~~-.-..-~~....-..-~ ............~
00 20.00 60.00 100.00
FREQUENCY, HZ
FIGURE 13 Transfer function at Point 1 (center of load)
at H = 20 ft-falling weight deflectometer. ..
Roesset and Shao 15

The computed deflections and the estimated moduli

C\J of the pavement, base, and subbase for the cases
ci studied are summarized in Table 3, Although the er-
rors are now much smaller, there are still some sig-
nificant differences for shallow soil deposits
POINT I (i.e., 10 ft) where the stiffness of the subbase is
overestimated and for some intermediate depths ( 40
ft) where the modulus of the subbase is underesti-
mated but that of the pavement is overestimated by
40 percent.

CONCLUSIONS

The results of the studies conducted to date indi-

cate that a static interpretation of the deflections
measured in the Dynaflect tests may be reasonable
when dealing with a homogeneous soil (subbase) ex-
POINT 7 tending to depths of 60 ft or more. When much
stiffer bedrock is encountered at shallower depths
important dynamic amplification can occur. The elas-
tic properties backfigured for the pavement system
using standard techniques can then be substantially
in error. The situation is aggravated when the soil
of the subbase is not homogeneous but its stiffness
increases with depth.
Dynamic effects are less important for the fall-
0.12 0.20 ing weight deflectometer because a broad range of
TIME (SEC) frequencies is excited instead of a single one, Even
so there are still some ranges of depth to bedrock
F1GURE 14 Time history of deflections at Points 1 and 7- for which the differences in dynamic effects at the
falling weight deflectometer. various stations (distorting the shape of the de-

0 0
0 0
v <t

IMP(H=20ft) STA .(H=INF.)

0 0
0 0
C\J C\J

0 0 +----.------.---.---.-----.-----l oo~- . . . . .--.----.----.------.-----l

00 2.00 4 .00 6.00 00 2.00 4 .00 6.00
Distance, ft Distance, ft

FIGURE 15 Ratio of dynamic (IMP) to static (H = ~) deflections at H = 20 ft-falling weight deflectometer.

TABLE 3 Deflection Bulbs and Estimated Elastic Moduli for Homogeneous

Subbase and Different Depths to Bedrock, Falling Weight Deflectometer
(displacement x 10- 8 ft)

Distance to the Center (ft)

Estimated Errors
H (ft) 0 2 4 6 (Jb/in 2 ) (%)

Static 11 .54 5.139 3. 141 2.180 1.611 1.253 1.015 200,000

78,500
29,000
Dynamic 10 10.60 4.622 2.842 1.9 23 1.31 7 0 .909 4 0.7214 200,000
78,500 0.0
35,539
Dynamic 20 11.06 4.652 3.013 2.073 1.538 1.280 1.090 200,000 0.0
82,200 4.7
28,790 0.7
Dynamic 40 10.74 4.860 3.008 2.111 1.590 l.288 1.086 287,200 43.6
87,375 11.3
28,331 2.3
Dynamic 80 11 .08 4.733 3.073 2.109 1.608 1.3 11 1.044 200,000 0.0
89,131 13.5
29,245 0.8
16 Transportation Research Record 1022

flection bulb) may lead to erroneous estimates of
the elastic moduli.
ACKNOWLEDGMENTS
The work described in this paper was conducted at
The University of Texas at Austin under a research
grant from the Texas State Department of Highways
and Public Transportation.
REFERENCES
1. W.T. Thomson. Transmission of Elastic waves
Through a Stratified Soil Medium. Journal of Ap-
plied Physics, Vol. 21, Feb. 1950.
2. N.A. Haskell. The Dispersion of Surface Waves on
Multilayered Media. Bulletin of the Seismologi-
cal Society of America, Vol. 43, No. 1, Feb .
1953.
3. E. Kausel and J.M. Roesset. Stiffness Matrices
for Layered Soils. Bulletin of the Seismological
Society of America, Vol. 71, No. 6, Dec. 1981.
4, G. Gazetas. Dynamic Stiffness Functions of Strip
and Rectangular Footings on Layered Soils. S.M.
thesis. Massachusetts Institute of Technology,
Cambridge, 1975.
5. R.J. Apsel. Dynamic Green's Functions for Lay-
ered Media and Applications to Boundary Value
Problems. Ph.D. dissertation. University of Cal-
ifornia, San Diego, 1979.
6. G. Waas. Linear Two-Dimensional Analysis of Soil
Dynamics Problems on Semi-Infinite Layered
Media. Ph.D. dissertation. University of Cal-
ifornia, Berkeley, 1972.
7. E, Kausel. Forced Vibrations of Circular Founda-
tions on Layered Media. Research Report R74-ll.
Department of Civil Engineering, Massachusetts
Institute of Technology, Cambridge, 1974.
8. E, Kausel. An Explicit Solution for the Green
Functions for Dynamic Loads in Layered Media.
Research Report RSl-13. Massachusetts Institute
of Technology, Cambridge, 1981.
9. w. Uddin, W.R. Meyer, and K.H. Stakoe. Project-
Level Structural Evaluation of Pavements Based
on Dynamic Deflections. !.!!_ Transportation Re-
search Record 1007, TRB, National Research Coun-
cil, Washington, n.c., 1985, pp. 37-45.
Publication of this paper sponsored by Committee on
Mechanics of Earth Masses and Layered Systems.

Pavement Evaluation Using Deflection Basin

Measurements and Layered Theory
ALBERT J. BUSH III and DON R. ALEXANDER

ABSTRACT

Recent developments through research efforts at the waterways Experiment Sta-

tion (WES) have produced a pavement evaluation procedure that uses deflection
basin measurements from nondestructive test devices. These deflections are in-
put for a layered elastic program (BISDEF) that predicts elastic moduli for
each pavement layer for up to a four-layer system. The approach has been veri-
fied through comparison of predicted moduli from the computer program to moduli
from laboratory modulus tests. The moduli determined from the deflection basin
and BISDEF are then used with limiting strain criteria and a layered elastic
program (AIRPAVE) to determine allowable aircraft loads, strengthening overlay
requirements, and so forth. The use of a single evaluation procedure that em-
ploys test results from six different nondestructive testing devices to deter-
mine the allowable aircraft load on flexible airfield pavements is evaluated.
Test data presented here were obtained from a side-by-side comparative study
conducted in October 1982 at MacDill Air Force Base on three different pave-
ments (two asphalt concrete and one composite of asphalt concrete over portland
cement concrete) , Test devices considered in this paper are the WES 16-kip
vibrator, three falling weight deflectometers, a Road Rater, and a Dynaflect.
Allowable loads determined using data from each device compare favorably with
the standard evaluation procedure. The moduli values for the base course mate-
rials are higher when a preload is applied as in the case of the WES 16-kip
vibrator.